Study On Aerodynamic Performance Prediction Of Wind Turbines Based On Similarity Theory

Predicting the aerodynamic performance of wind turbine is a key point to the wind turbine design. Accurately predicting it will be beneficial for the blade aerodynamic design, blade structural design, system evaluation and so on. With the development of offshore wind power, the size of wind turbine is becoming larger and larger, which induces more complicated flow around the longer blade and leads to the more difficult prediction of aerodynamic performance. However, the precision of aerodynamic performance prediction with existing methods including BEM and CFD can’t meet the requirements. Wind farm test and wind tunnel test to wind turbine can’t also achieve the goal, because there are a lot of uncontrolled factors during wind farm test, and the size of test model is limited by wind tunnel size so that it can’t reach the same Reynolds number of real wind turbine. Therefore, it is significant to carry out a study of the wind turbine aerodynamic performance prediction based on similarity theory in order to fast and accurately evaluate the large size wind turbine aerodynamic characteristics.In this thesis, the study was carried out with CFD calculations and wind tunnel experiments based on similarity theory.Firstly, the theoretical similarity criteria and similitude function relationship of wind turbine aerodynamic performance have been obtained by using dimensional analysis and equation analysis of similarity theory. The three theoretical similarity criteria are relative surface roughness criterion, Reynolds number criterion and tip speed ratio criterion, and the theoretical similitude function relationship is CP/CM/CT=/(ε, Rec, λ). A successive approximation method which considers the part of similarity criteria was proposed.Secondly, the aerodynamic characteristics and flow behaviors of NREL Phase VI wind turbine (diameter=10.058m) and its two scaled models (diameter=1.5m and1.0m) with smooth blades were investigated by using CFD software. It was found that their power coefficients (Cp) and axial thrust coefficients (CT) were different at the same Reynolds number (Rec(0.75R)≤7.42×105) and tip speed ratio (λ≤10.00), which showed that the aerodynamic performances among the models didn’t agree with the theoretical similarity law. The reasons causing the differences on the aerodynamic performances were revealed by comparing and analyzing the flow information of models.Thirdly, the problem that roughness on blade surface is difficult to be completely similar between the wind turbine and its scaled models is inevitable. Therefore, it needs to investigate the effects of surface roughness on the aerodynamic performance of wind turbine blade and design the roughness insensitivity blade so that the relative surface roughness principle mentioned above can be neglected in the course of aerodynamic performance prediction. In this thesis, the evaluation indicators for roughness sensitivity on wind turbine blade were obtained by analyzing the data of real variable-speed and pitch-regulated wind turbines with BEM theory. Moreover, the method that the evaluation benchmark was determined by determining the weight coefficients of the evaluation indicators for roughness sensitivity with the impacts of lift and drag coefficients on power output was proposed. At the same time, the effects of21%relative thickness airfoil geometry parameters on the evaluation indicators for roughness sensitivity were also investigated by using orthogonal design and analysis of variance as an example for the design of the roughness insensitivity wind turbine blade. In addition, the roughness sensitivity of CAS thin airfoils was evaluated by comparing with the data obtained in wind tunnel test and from XFOIL calculation. And a new CAS-W2-210airfoil which had lower roughness sensitivity and better aerodynamic performance than the original CAS-W1-210was obtained by modifying the geometry parameters.Fourthly, it is necessary to recognize the aerodynamic characteristics of airfoils at low Reynolds numbers during the study of wind turbine similarity, because the airfoils of a scaled model blade usually run within the low Reynolds number range. In this thesis, S809airfoil was tested in wind tunnel at less than Re=5×105and obtained a series of aerodynamic performance data which filled the data gaps. The results showed that the aerodynamic characteristics of S809airfoil were complex at those low Reynolds numbers. At the same time, the experiments about the airfoil aerodynamic performances influenced by wing tip vortex were carried out at low Reynolds numbers, in order to further study the influences of blade tip vortex on the blades of scaled models. The results about the pressure lift and drag coefficients under the effects of wing tip vortex were obtained at different Reynolds numbers. In addition, the experiments about the rough airfoil aerodynamic performances influenced by wing tip vortex were also done at low Reynolds numbers in order to further study on the roughness sensitivity of the scaled model blades under the effects of blade tip vortex. Based on the above experiments, the results about the pressure lift and drag coefficients of the rough and clean airfoils under the effects of wing tip vortex, the changes of vortex core position and vorticity with the angle of attack and airfoil surface status were obtained at different Reynolds numbers.Finally, the correction for the aerodynamic performances of scaled models was investigated. It showed that the result of Prandtl tip loss correction didn’t well agree with the2D airfoil data tested in wind tunnel under the attached flow condition at low Reynolds numbers. It also showed that there were great differences between the3D data corrected by existing3D correction methods based on2D airfoil data tested in wind tunnel and the3D CFD results. Although Du and Selig method had a least error among those methods, it needs to be further improved for its application. On the other hand, the correction for the flow similarity of scaled models by correcting inflow velocity was also investigated. Their stream tubes, stream lines and velocity distributions were not similar because of their different flow losses in flow fields, when the flow rates were similar between prototype and its scaled models. Therefore, the flow rate and flow loss should be two factors of the correction for their flow similarity.